[GIT PULL] please pull infiniband.git

[GIT PULL] please pull infiniband.git

[Roland Dreier]

Linus, please pull from

    master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband.git for-linus

This tree is also available from kernel.org mirrors at:

    git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband.git for-linus

This has a few more fixes for 2.6.31-rc1; however everything here
should be a fix appropriate for -rc2 as well, so if this misses the
end of the merge window then I think it should still be OK to pull.

Alexander Schmidt (1):
      IB/ehca: Bump version number

Faisal Latif (2):
      RDMA/nes: Fix max_qp_init_rd_atom returned from query device
      RDMA/nes: Fix FIN state handling under error conditions

Hannes Hering (1):
      IB/ehca: Tolerate dynamic memory operations before driver load

Peter Huewe (1):
      RDMA: Add __init/__exit macros to addr.c and cma.c

Roel Kluin (1):
      IB/ehca: Ensure that guid_entry index is not negative

Roland Dreier (3):
      IB/mthca: Replace dma_sync_single() use with proper functions
      mlx4_core: Fix dma_sync_single_for_cpu() with matching for_device() calls
      Merge branches 'ehca', 'misc', 'mlx4', 'mthca' and 'nes' into for-linus

 drivers/infiniband/core/addr.c         |    4 +-
 drivers/infiniband/core/cma.c          |    4 +-
 drivers/infiniband/hw/ehca/ehca_hca.c  |    2 +-
 drivers/infiniband/hw/ehca/ehca_main.c |   20 +-
 drivers/infiniband/hw/ehca/ehca_mrmw.c |  508 +++++++++++++++++++++++++++++++-
 drivers/infiniband/hw/ehca/ehca_mrmw.h |   13 +-
 drivers/infiniband/hw/mthca/mthca_mr.c |   13 +-
 drivers/infiniband/hw/nes/nes_cm.c     |    8 +-
 drivers/infiniband/hw/nes/nes_verbs.c  |    2 +-
 drivers/net/mlx4/mr.c                  |   14 +-
 10 files changed, 554 insertions(+), 34 deletions(-)

Re: [ofa-general] [GIT PULL] please pull infiniband.git

[Tziporet Koren]

Roland Dreier wrote:
> Linus, please pull from
>
>     master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband.git for-linus
>
> This tree is also available from kernel.org mirrors at:
>
>     git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband.git for-linus
>
> This has a few more fixes for 2.6.31-rc1; however everything here
> should be a fix appropriate for -rc2 as well, so if this misses the
> end of the merge window then I think it should still be OK to pull.
>
>   

Roland
What about the patches from Yevgeny supporting our new ConnextX HW 
coming soon?

I talk about:
http://lists.openfabrics.org/pipermail/general/2009-June/060225.html
http://lists.openfabrics.org/pipermail/general/2009-June/060226.html

Since device ID is already in pci.ids I think it is important to have 
this code too, otherwise the driver will not load :-(

Thanks
Tziporet

Re: [ofa-general] [GIT PULL] please pull infiniband.git

[Roland Dreier]

 > What about the patches from Yevgeny supporting our new ConnextX HW
 > coming soon?
 > 
 > I talk about:
 > http://lists.openfabrics.org/pipermail/general/2009-June/060225.html
 > http://lists.openfabrics.org/pipermail/general/2009-June/060226.html

Sorry, missed that going by the first time, will reply to the messages.

Re: [ofa-general] [GIT PULL] please pull infiniband.git

[Tziporet Koren]

Roland Dreier wrote:
> Linus, please pull from
>
>     master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband.git for-linus
>
> This tree is also available from kernel.org mirrors at:
>
>     git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband.git for-linus
>
> This will get a second batch of InfiniBand/RDMA batches.  In addition
> to the usual motley crew of changes, this pull includes a new driver
> for NetEffect RNICs in drivers/infiniband/hw/nes.  The code could use
> some further cleaning, but I don't think it's worth holding off on the
> merge.
>
> David Dillow (1):
>       IB/srp: Retry stale connections
>
> Eli Cohen (2):
>       IB/mthca: Remove checks for srq->first_free < 0
>       IB/ib_mthca: Pre-link receive WQEs in Tavor mode
>
> Glenn Streiff (1):
>       RDMA/nes: Add a driver for NetEffect RNICs
>
> Hoang-Nam Nguyen (1):
>       IB/ehca: Add PMA support
>
> Jack Morgenstein (2):
>       IB/mthca: Don't read reserved fields in mthca_QUERY_ADAPTER()
>       mlx4_core: Don't read reserved fields in mlx4_QUERY_ADAPTER()
>
> Joachim Fenkes (2):
>       IB/ehca: Prevent sending UD packets to QP0
>       IB/ehca: Update sma_attr also in case of disruptive config change
>
> Olaf Kirch (1):
>       IB/mthca: Return proper error codes from mthca_fmr_alloc()
>
> Or Gerlitz (3):
>       IPoIB: Handle bonding failover race for connected neighbours too
>       IPoIB: Remove a misleading debug print
>       IB/fmr_pool: Allocate page list for pool FMRs only when caching enabled
>
> Roland Dreier (4):
>       mlx4_core: Fix more section mismatches
>       IB/mthca: Fix and simplify page size calculation in mthca_reg_phys_mr()
>       IB/mlx4: Actually print out the driver version
>       IB: Avoid marking __devinitdata as const
>
> Sean Hefty (1):
>       IB/cm: Add interim support for routed paths
>
>
>   

Does this means ConnectX WQE BB and stateless offload will not be in for 
2.6.25?

Tziporet

Re: [ofa-general] [GIT PULL] please pull infiniband.git

[Jack Morgenstein]

On Tuesday 27 November 2007 08:21, Roland Dreier wrote:
> Linus, please pull from
> 
>     master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband.git for-linus
> 
> 
> Jack Morgenstein (1):
>       mlx4_core: Fix state check in mlx4_qp_modify()
> 
MUST also enter the patch I send yesterday:
[PATCH] ipoib: fix kernel Oops resulting from xmit when priv->broadcast is NULL.

(critical bug fix -- will get kernel Oopses whenever ports on the network go down!).
(patch given again below)

- Jack
======================================================================================================================
IPoIB: Fix kernel Oops resulting from xmit following dev_down.

If a port goes down, ipoib_ib_dev_down is invoked -- which flushed the mcasts (clearing
priv->broadcast) and clearing the path record cache. If ipoib_start_xmit is then invoked (before
the port is upped), a kernel Oops results from attempting to access priv->broadcast.

Returning NULL if priv->broadcast is NULL is a harmless way of bypassing the problem -- the
offending packet is simply discarded "without prejudice".

Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>

diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c
index a03a65e..c9f6077 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_main.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c
@@ -460,6 +460,9 @@ static struct ipoib_path *path_rec_create(struct net_device *dev, void *gid)
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ipoib_path *path;
 
+	if (!priv->broadcast)
+		return NULL;
+
 	path = kzalloc(sizeof *path, GFP_ATOMIC);
 	if (!path)
 		return NULL;